7/23/2019 Semi-Automatic Classification Plugin- Tutorial http://slidepdf.com/reader/full/semi-automatic-classification-plugin-tutorial 1/20 12/9/2015 3. Tutor ial s — Sem i- Autom ati c C lassi fi cati on Pl ugi n 2.5.1 docum entati on http://semiautomaticclassificationmanual.readthedocs.org/en/latest/Tutorials.html 1/20 3. Tutorials The following is a basic tutorial about the use of the Semi-Automatic Classification Plugin. However, visit the blog From GIS to Remote Sensing for new and updated tutorials such as: Estimation of Land Surface Temperature with Landsat Thermal Infrared Band; Land Cover Classification of Cropland. In this tutorial we are going to classify a Landsat image (single band rasters). Download the sample dataset, which is a Landsat 8 image (a subset acquired in the South of Rome, Italy) available from the U.S. Geological Survey. The following bands (each band is a single 16 bit raster) are included in the file: 2 - Blue; 3 - Green; 4 - Red; 5 - Near-Infrared; 6 - Short Wavelength Infrared 1; 7 - Short Wavelength Infrared 2. The Semi-Automatic Classification Plugin uses SAGA GIS for the classification process. The SAGA algorithms work only with single band images as input. Therefore, if input is a multi band raster (i.e. a single image file made of several bands), the Semi-Automatic Classification Plugin automatically splits the input file to single band rasters, which takes some time depending on raster size. In order to optimize the classification process (especially for hyperspectral images), it is preferable to use single band rasters, or split the image file to single bands, as explained here (point 1) . 3.1. Define the inputs Required inputs are a multi band raster or single band rasters, and a training shapefile for ROI creation. In this tutorial we are going to create a band set. Start QGIS and load the raster bands; create a color composite (RGB=543) of the sample data, which is useful for the photo interpretation, as explained here (point 2); start the Semi-Automatic Classification Plugin (if the docks of the are not displayed, click the button Show docks in the main plugin interface);
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
The following is a basic tutorial about the use of the Semi-Automatic Classification
Plugin. However, visit the blog From GIS to Remote Sensing for new and updated
tutorials such as:
Estimation of Land Surface Temperature with Landsat Thermal Infrared Band;
Land Cover Classification of Cropland.
In this tutorial we are going to classify a Landsat image (single band rasters). Download
the sample dataset, which is a Landsat 8 image (a subset acquired in the South of Rome,
Italy) available from the U.S. Geological Survey. The following bands (each band is a
single 16 bit raster) are included in the file:
2 - Blue;
3 - Green;
4 - Red;
5 - Near-Infrared;
6 - Short Wavelength Infrared 1;
7 - Short Wavelength Infrared 2.
The Semi-Automatic Classification Plugin uses SAGA GIS for the classification process.
The SAGA algorithms work only with single band images as input. Therefore, if input is amulti band raster (i.e. a single image file made of several bands), the Semi-Automatic
Classification Plugin automatically splits the input file to single band rasters, which takes
some time depending on raster size. In order to optimize the classification process
(especially for hyperspectral images), it is preferable to use single band rasters, or split
the image file to single bands, as explained here (point 1) .
3.1. Define the inputs
Required inputs are a multi band raster or single band rasters, and a training shapefile
for ROI creation. In this tutorial we are going to create a band set.
Start QGIS and load the raster bands; create a color composite (RGB=543) of the
sample data, which is useful for the photo interpretation, as explained here (point
2); start the Semi-Automatic Classification Plugin (if the docks of the are not
displayed, click the button Show docks in the main plugin interface);
If the checkbox Calculate accuracy is checked, than the error matrix is calculated
and saved as file .txt in the same directory of the .tif file (also, it is automatically
displayed in the tab of the plugin main interface Post processing > Accuracy); the
error matrix is calculated by comparing the classification to the training shapefile
used for the classification (see below Post processing tools);
It is possible to apply a mask shapefile to the classification; download this
shapefile , check the checkbox Apply mask and select the downloaded shapefile;click the button Perform classification, and the classification will be saved along
It is possible to assess the classification accuracy (implemented by GRASS GIS),
by comparing the classification to a reference shapefile (not necessarily the
training shapefile); now, select the tab Post processing > Accuracy of the plugin
main interface; select theclassification.tif beside Select a classification to assess
and select the ROI shapefile beside Select the reference shapefile; then click thebutton Calculate error matrix and the matrix will be displayed; you can save the
error matrix by clicking the button Save error matrix to file;
It is useful to calculate the land cover change (through GDAL and Numpy)
between a reference classification raster and a new classification raster; download
this classification (pretend this is the last year classification); select the tab Postprocessing > Land cover change of the plugin main interface, select the
downloaded classification as reference classification, and the classification.tif as
the new classification; click the button Calculate land cover change and select
where to save the raster of changes and the related table (i.e. a file .csv, whose
values are separated by tab); pixel values of the raster of changes (ChangeCode)
are described in the table, and each value represent a class of change from the
reference classification to the new classification;